Hi guys, I would like to share a side project I was working on to help cyclists increase chances to win races on Zwift.
ZRace analyzes all athletes registered in a Zwift race and predicts possible winners. It also analyzes each category and presents the average power required for you to have a good result. In addition, athletes with specific profiles are identified, such as climber, sprinter, and time-trialist.
If you like racing, virtually or not, you will probably enjoy using the tool or reading the article on how I developed it.
Article: Predicting winners in cycling races with Machine Learning | by Bruno Gregory | Mar, 2021 | Medium
Video from GPLama:
I hope you guys like it. Please feel free to comment, suggest changes and features. Any feedback is welcome.
Very cool, subbing to thread to see how this progresses. Great work, thanks and will give it a shot when the weather turns poor for a bit!
How is it working for you in practise accuracy wise, any success yet?
Wow, thanks for sharing your research and results. Your article was a great read.
Once Trainerroad and Best Bike Split merge it will be cool to see your work added to the solution to create the definitive race preparation and race day “operating system” for racing cyclists and coaches.
It’s cool having a crystal ball…
Wait I see another vision, Trainerroad partnering with Head Space to create cycling specific content.
In practice, the tool really helps especially to analyze which race you can have a better result. For example, you can understand that you have no condition to be in a category A race.
Hello Piller, I’m glad you liked the article. I also liked your vision.
I think folks from TR will also like it, especially @Nate_Pearson , who is just like me, a data-driven guy.
I guess if you’re accessing things like https://zwiftpower.com/cache3/results/906655_zwift.json there’s a chance Zwift might shut you out, but I hope they don’t.
Maybe from a GDPR angle, as people who have opted in have agreed to have their data processed by ZwiftPower; that doesn’t extend to third parties.
It’s really interesting work though, so with any luck no one will notice.
Hi Derenc, yes everything you said has already been mentioned in the video.
I added a new feature to the tool. It is now possible to check the statistical data of the races, riders, and other information.
I particularly find these data very relevant whether you race virtually or not. The profile data of the winners by category should reflect well what we see in non-virtual races too.
Hope you like it.
Nice work! I can, with 100% accuracy, predict the loser of any Zwift race I participate in.